Molecular Phylogenetics and Evolution 106 (2017) 228–240 Contents lists available at ScienceDirect Molecular Phylogenetics and Evolution journal homepage: www.elsevier.com/locate/ympev Resolving incongruence: Species of hybrid origin in Columnea (Gesneriaceae) ⇑ James F. Smith a, , John L. Clark b,c, Marisol Amaya-Márquez d, Oscar H. Marín-Gómez d,e a Department of Biological Sciences, Boise State University, 1910 University Drive, Boise, ID 83725-1515, USA b Department of Biological Sciences, The University of Alabama, Box 870345, Tuscaloosa, AL 35487, USA c The Lawrenceville School: Science Department, The Lawrenceville School, 2500 Main Street, Lawrenceville, NJ 08648, USA d Instituto de Ciencias Naturales, Universidad Nacional de Colombia, Apartado 7495, Bogotá, Colombia e Posgrado en Ciencias, Instituto de Ecología, INECOL A.C, Carretera antigua a Coatepec 351, El Haya, Xalapa 91070, Veracruz, Mexico article info abstract Article history: Speciation by hybridization has long been recognized among plants and includes both homoploid and Received 12 August 2016 allopolyploid speciation. The numbers of presumed hybrid species averages close to 11% and tends to Revised 30 September 2016 be concentrated in a subset of angiosperm families. Recent advances in molecular methods have verified Accepted 3 October 2016 species of hybrid origin that had been presumed on the basis of morphology and have identified species Available online 5 October 2016 that were not initially considered hybrids. Identifying species of hybrid origin is often a challenge and typically based on intermediate morphology, or discrepancies between molecular datasets. Keywords: Discrepancies between data partitions may result from several factors including poor support, incom- Hybrid species plete lineage sorting, or hybridization. A phylogenetic analysis of species in Columnea (Gesneriaceae) Interspecific hybrids nrDNA capture indicated significant incongruencies between the cpDNA and nrDNA datasets. Tests that examined whether one or both of the datasets had the phylogenetic signal to reject the topology of the alternate dataset (Shimodaira and Hasegawa [SH] and approximately unbiased [AU] tests) indicated significant dif- ferences between the topologies. Splitstree analyses also showed that there was support for the place- ment of the discrepant taxa in both datasets and that the combined data placed the putative hybrid species in an intermediate position between the two datasets. The genealogical sorting index (GSI) implied that coalescence in nrDNA had occurred in all species where more than a single individual had been sampled, but the GSI value was lower for the cpDNA of most of the putative hybrids, implying that these regions have not yet coalesced in these lineages despite being haploid. The JML test that eval- uates simulated species pairwise distances against observed distances also implies that observed nrDNA data generate shorter distances than simulated data, implying hybridization. It is most likely that C. gigantifolia, C. rubriacuta, and C. sp. nov. represent a lineage from a hybrid ancestor, but C. moorei may be a more recent hybrid and may still be undergoing hybridization with sympatric species. Ó 2016 Elsevier Inc. All rights reserved. 1. Introduction and founder effects (Glor et al., 2005; Smith et al., 2014; Schulte et al., 2015). More rapid means of generating species include The process of speciation has long been a focus of evolutionary hybridization either through homoploid or polyploid means and and systematic biology (Darwin, 1859; Mayr, 1942; Levin, 1978, generally is estimated to be responsible for 11% of all species 2000; Coyne and Orr, 2004; Lexer and Widmer, 2008; Givnish, (Stebbins, 1959; Rieseberg et al., 1996a; Rieseberg, 1997, 2006; 2010, 2015; Sochar et al., 2015; Kadereit, 2015; Arnold, 2016). Wendel and Cronn, 2003; Soltis and Soltis, 2009; Abbott et al., Allopatric speciation is a well documented process and often is 2013). Detecting hybrids is not always a simple task. Intermediacy divided into vicariant speciation where landscapes change to iso- in morphology has long been a staple for detecting hybrids, but it is late populations from each other (Pereira and Baker, 2004; now clear that hybrids may possess transitive morphologies: char- Struwe et al., 2009; Bentley et al., 2014), or isolation by dispersal acter states that are unknown in the parental species (Rieseberg and Ellstrand, 1993). For example, Castilleja christii N. H. Holmgren was recently documented to be a homoploid hybrid species (Clay ⇑ Corresponding author. et al., 2012), but was never suspected to be of hybrid origin. In E-mail address: [email protected] (J.F. Smith). http://dx.doi.org/10.1016/j.ympev.2016.10.001 1055-7903/Ó 2016 Elsevier Inc. All rights reserved. J.F. Smith et al. / Molecular Phylogenetics and Evolution 106 (2017) 228–240 229 large part the lack of suspicion stemmed from C. christii having pri- Aeschynanthus (Denduangboripant and Cronk, 2000). Therefore marily yellow bracts in contrast to the crimson and scarlet bracts we evaluate whether incomplete lineage sorting or hybridization of the parental species, C. miniata and C. linariifolia. Castilleja christii may be most prevalent. We examined our data using several alter- was shown to be a hybrid only when direct sequencing of low copy native analytical approaches including using Shimodaira and Hase- nuclear genes demonstrated multiple peaks in chromatograms that gawa (SH; Shimodaira and Hasegawa, 1999) and approximately corresponded to the two separate peaks that were present in the unbiased (AU; Shimodaira, 2002) tests and two coalescent parental copies of the homologous gene (Clay et al., 2012). approaches. The first used the genealogical sorting index (GSI, Modern means of detecting hybrids often occur when a species Cummings et al., 2008) that evaluates whether a pre-defined clade is placed in different clades when chloroplast and nuclear DNA is monophyletic even when it may not be recovered as mono- data are analyzed independently (Smith and Sytsma, 1990; phyletic in standard phylogenetic analyses. Clades that are mono- Rieseberg and Soltis, 1991; Rieseberg et al., 1996a or b; Baum phyletic are more likely to have achieved coalescence and et al., 1998; Wendel and Doyle, 1998; Linder and Rieseberg, therefore are less likely to reflect incomplete lineage sorting 2004; Howarth and Baum, 2005; Friar et al., 2008; Rothfels et al., (Palumbi et al., 2001; Hedrick, 2007; de Viliers et al., 2013). 2015; Walker et al., 2015). Discrepancies between data partitions Joly et al. (2009) and Joly (2012) developed the software JML to have posed challenges to phylogenetic analyses since systematists detect whether intraspecific variation in a gene that produces a started comparing more than a single dataset (Kluge, 1989; Smith discrepant relationship relative to another gene is likely the result and Sytsma, 1994; Mason-Gamer and Kellogg, 1996; Smith, 2000). of incomplete coalescence. If incomplete lineage sorting can be Incongruences can occur from a multitude of causes and the great- eliminated, the probability that the incongruency is the result of est challenge is to resolve the source of the incongruency. One pos- hybridization is increased. The software uses a posterior distribu- sibility is that the incongruence is the result of poor support in one tion of species trees, simulates gene trees and DNA sequences, then or more of the partitions (Farris et al., 1994; Seelanan et al., 1997; calculates the minimum distance between simulated sequences for Morrison, 2009). As a result, random noise and homoplasy may all pairs of species. These are then compared to the empirical data. have as much influence on the resulting topology as the phyloge- If the observed distances are smaller than the simulated distances, netic signal in the data. Such discrepancies are largely overcome then hybridization is a better explanation to account for the more with more data (sometimes by combining several datasets with recent common ancestry of the sequences than incomplete lineage weak, but essentially congruent signal (Smith, 2000) or data that sorting. has a higher proportion of phylogenetically informative characters The taxonomic focus of the present study is the genus Columnea (Small et al., 1998). In other cases the datasets may each strongly (Gesneriaceae). Columnea is a genus of over 209 species that has support conflicting relationships. In such cases, the challenge is been the focus of several recent phylogenetic investigations to to discover the biological explanation for the incongruency and re-evaluate the subgeneric classification system and to explore requires knowledge beyond what is recovered in the phylogenetic character state evolution (Smith et al., 2013; Schulte et al., 2014, analyses. In cases where there is potential for paralogy such as low 2015). Preliminary analyses that included two individuals of C. copy nuclear genes, the inclusion of non-orthologous loci for some rubriacuta (Wiehler) L.P. Kvist & L.E. Skog, and one cultivated indi- taxa will generate incongruency (Rokas et al., 2003). In these cases vidual of C. moorei C.V. Morton, had indicated these species were removing the paralogs may provide a simple answer as long as incongruently placed in the phylogeny using cpDNA or nrDNA. resolving which sequences are the ones in conflict can be identified They were excluded from the analyses pending increased sampling (Schulte et al., 2015). In other cases, incongruencies can occur with